I am only interested in single subject results in this case, although we do have another run for this subject. I have normalized the data. I overlaid the activation onto the neuroantomical image and it was outside the brain. In the glass brain, it was inside the brain. So, I think I will try following the advice of a previous post.
ref.img: T1 structural image
sou.img: Mean fMRI
other images: all the realigned fMRI volumes
From: Jonathan Peelle [mailto:[log in to unmask]]
Sent: Friday, May 28, 2010 9:47 AM
To: Modestino, Edward J *HS
Cc: [log in to unmask]
Subject: Re: [SPM] Coregistration/activity outside occipital lobe
> When I tried to overlay the functional activity onto the anatomical image, there is clear activity outside the occipital lobe. The image realignment revealed no greater than 1.1 mm maximal out of all three planes. What I found strange was the coregistration step. In my memory, I thought that we coregistered the functional images to the anatomical image. The anatomical image was a template. However, following the SPM manual and example, it was suggested to use the mean functional image was to be used as a template to which the anatomical image was to be aligned. This is what I did. Could this be the problem? Should I have coregistered the functional images to the anatomical (as a template)? Thus, the functional activation would be forced into the anatomical space and when overlain, would be inside the brain.
At this point, are you interested in only single-subject results, or
group statistics? I ask because you didn't mention anything about
normalization to MNI space.
In the case where you haven't normalized, if you're coregistered the
mean functional (and thus all of the realigned functional images) and
the structural, then your stats and structural image should be in the
same space. [Note that when coregistering, you would typically enter
the mean functional as the reference image, and the source image as
the structural. This means that the structural is moved to match the
mean functional (and thus all of the other functional images as well).
If you do it the other way around, the mean functional and structural
will be aligned, but they will not be aligned with all of the other
functional images (because you've moved the mean functional, but not
the other functional images), and thus your stats will not line up
with the structural image (unless you enter all the other functional
images in "other images" during coregistration---but easier to just
have mean functional as reference, structural as source).] So this is
one potential cause.
Second, assuming your stats line up with your structural images, at
this point they won't be in MNI space. So if you run results and look
at the MIP (glass brain) rendering, things might appear outside the
brain because the brain outline is really only for normalized images.
However, if you go to overlays > sections and select the subject's
structural image, things should line up ok.
Finally, it might be a good idea to normalize to MNI space anyway,
either for group statistics or just to report MNI coordinates of your
results. In which case, following coregistration, you can segment
your structural image, which will produce parameters (*seg_sn.mat) for
warping to MNI space. You can then use the "normalise: write" button
to apply these parameters to all of your (coregistered) functional
Hope this helps!